Hypergraph Modelling and Graph Clustering Process Applied to Co-word Analysis

Xavier Polanco (1, 2) , Eric Sanjuan (3)

1 MALIRE - Machine Learning and Information Retrieval
2 INIST - Institut de l'information scientifique et technique
3 LIA - Laboratoire Informatique d'Avignon

We argue that any document set can be modelled as a hypergraph, and we apply a graph clustering process as a way of analysis. A variant of the single link clustering is presented, and we assert that it is better suited to extract interesting clusters formed along easily interpretable paths of associated items than algorithms based on detecting high density regions. We propose a methodology that involves the extraction of similarity graphs from the indexed-dataset represented as a hypergraph. The mining of informative short paths or geodesics in the graphs follows a graph reduction process. An application for testing this methodology is briefly exposed. We close this paper indicating the future work.

Mots clés: Co-word analysis; Hypergraph modelling; Graph clustering; Minimal traversal; Formal concept; Association rules; Text mining

Domaines: Intelligence artificielle [cs.AI]; Mathématique discrète [cs.DM]

Xavier Polanco, Eric Sanjuan. Hypergraph Modelling and Graph Clustering Process Applied to Co-word Analysis. ISSI 2007 - 11th International Conference of the International Society for Scientometrics and Informetrics, Jun 2007, Madrid, Spain. pp.613-618. ⟨hal-00165984⟩